Detection of the Geothermal Alterations and Thermal Anomalies by Processing of Remote Sensing Data, Sabalan, Iran
Total Page:16
File Type:pdf, Size:1020Kb
DETECTION OF THE GEOTHERMAL ALTERATIONS AND THERMAL ANOMALIES BY PROCESSING OF REMOTE SENSING DATA, SABALAN, IRAN Tohid NOURI, Majid Mohammady Oskouei Sahand University of Technology, Tabriz, Iran, Tel:+984123459299 Email: [email protected], [email protected] KEYWORDS: ASTER, linear spectral unmixing, geothermal, hydrothermal alteration, thermal anomaly. Abstract: Geothermal energy is directly related to radioactive elements inside of the earth. In geothermal areas thermal gradient is higher than vicinity regions. Extraction of thermal energy is therefore more beneficial in these areas. The Sabalan, Damavand, Sahand, Taftan, and Bazman regions have been proved as good potentials for geothermal energy in Iran. This study aims assessment of capability of remote sensing technology for detection of geothermal resources in Sabalan geothermal area. To do this task, processing of two datasets of Advanced Spaceborn Thermal Emission and Reflection radiometer (ASTER) data were used for alteration detection. Two ASTER scenes from the study area were therefore merged and processed. The applied method for mineral detection in this research is linear spectral unmixing. The important alterations (carbonate, borate, iron oxides, and clay minerals) were successfully detected and mapped. Besides that, the thermal anomalies were also investigated and mapped with the use of ASTER thermal channels. The resultant maps were then validated with comparing to the results of geophysical surveys available from SUNA organization. According to their works, the anomaly maps resulted from ASTAR data processing are able to illuminate the geothermal potential in the study area. INTRODUCTION Geothermal energy originates from collapsing of existing radioactive elements inside of the earth. In geothermal areas this source of heat energy, in the form of magma, is closer to the surface than other places. This kind of energy is one of the clean energy types and could be used directly for electricity generating. Detection of geothermal resources, therefore, has been much of interest for earth scientists. Considering the alteration caused by magma and hydrothermal fluids in surrounding rocks particularly around faults and fractures, a diversity of geochemical methods have been applied to geothermal resources explorations. Sinters, the chemical precipitates of hydrothermal systems, generally consist of mineral dominated by silica, carbonate, borates, metallic sulfides and oxides, and clay minerals (Hellman etal. 2004, Coolbaugh etal. 2006, Kratt etal. 2006). Therefore, areas that show anomalies of above mentioned minerals will be good targets for planning more detail exploration. Detection of such regions can be easily done by processing of remote sensing data. Exploration of Geothermal energy with remote sensing technology is useful in the early stages of exploration (Fernández etal. 2001). For a long time, geological remote sensing researchers have focused on the use of spectral signatures for rock type discrimination and mineral mapping, especially hydrothermal alteration minerals (Rowan etal. 2003). The use of remote sensing data analysis for geothermal exploration has been investigated in previous researches (e.g Coolbaugh etal. 2006,2007, Kratt etal. 2006,2009,2010, Eneva etal. 2007,2009). Recent developments in processing of both multi and hyperspectral data have led to the extensive application of those data in mineral detection. ASTER is one of the spaceborn multispectral satellite imagery systems that have been used for mineral exploration and it has better spectral resolution both in SWIR and thermal regions comparing to LANDSAT imagery (Azizi etal. 2010). As a result, on the basis of spectral characteristics of minerals, different alteration minerals are detected and mapped using ASTER data (Tangestani etal. 2008). Because of vicinity of geothermal systems to land surface, detection of geothermal related thermal anomalies could be another technique for underground geothermal systems exploration and thermal channels of ASTER data are also applied for surface temperature assessments and analysis. In this study, surface indicators (alteration minerals and thermal anomalies) of geothermal resources in Sabalan Mountain were detected. To do this task, the clay minerals, silica, borates, carbonates, and iron oxides were selected as indicator. Spectral image processing techniques including endmembers detection and unmixing algorithm were performed on the ASTER data. Finally we processed thermal bands of ASTER for extraction of geothermal related surface thermal anomalies. DATA ACQUISITION AND PREPROCESSING In this study two georeferenced scenes of ASTER level 1B data were merged to achieve necessary coverage of the region. In figure 1 time of acquisition and position of data is illustrated. The data was then atmospherically and topographically corrected using FLAASH and Lambertian (Riaño etal. 2003) method respectively before running the unmixing and thermal procedures. Figure 1: position an d time of acquisition the ASTER data. LINEAR SPECTRAL UNMIXING ALGORITHM (LSU) In this study geothermal related minerals were detected using unmixing algorithm. In LSU algorithm, as indicated in equation 1, an unknown pixel usually consisted of different materials and the total reflection of a pixel is considered as a linear mixings of reflections of the materials (Chang. 2007, Borengasser etal. 2008) where X is pixel spectrum, is coefficient of ith reference spectrum (endmember), is ith reference spectrum, and is noise. Unmixing algorithm determines contribution of different materials in all pixels of data with decomposition of their spectrums. Before implementation of Unmixing algorithm four steps needs to be done including: 1) minimum noise fraction (MNF), 2) pixel purity index (PPI), 3) n-dimensional visualization (n- DV), 4) spectral analysis. (1) MINIMUM NOISE FRACTION (MNF) MNF is statistical method similar to PCA (Azizi etal. 2010) that separates noise from data and gives estimation about actual dimensionality of data and reduces latter computations. After preprocessing, the MNF algorithm was implemented for noise whitening and preparing the data for Pixel Purity Index (PPI). In the MNF derived image there is no correlation among bands and its first band reflects main part of information, as indicated in the figure 2 due to its higher eigenvalue comparing to following bands (figure 3). Figure 2: MNF image of the study area. (a) band. 1 and (b) band. 2. Figure 3: Eigenvalues of MNF image bands 1 to 9 PIXEL PURITY INDEX (PPI) Recently many algorithms were developed for extraction of pure pixels of remote sensing multi and hyperspectral data and PPI is one of efficient algorithms for this task. PPI is the method that determines relative purity of pixels using the convex geometry argument (Qiu etal. 2006). In the PPI algorithm n random unit vectors are generated in m-dimensional data space (m is the number of MNF image bands in this case) and all pixels are projected to these random vectors. After predefined iterations, pixels that their projections fall far from mean projection by a certain threshold are marked as pure pixels. PPI algorithm was therefore implemented over MNF image for detection of purest pixels. Most of the pure pixels of the study area are in the western and eastern parts as indicated in figure 4. N-DIMENSIONAL VISUALIZATION In the PPI algorithm pixels are evaluated in terms of purity and characteristics of pure pixels are not recognized. They were plotted in an n-dimensional space for grouping of pure pixels. This is done by ENVIs N- Dimensional Visulizer and 5 classes were detected by visual interpretation (Figure 5). For determination of mineralogy of detected classes the mean spectrum of each of them were calculated. Figure 4: Pure pixels of the Figure 5: clouds of pure pixels in 3D view and 5 distinguished classes. study area (red pixels). SPECTRAL ANALYSIS Spectral analysis is used to identify different mineral types based on their spectral features (Qiu etal. 2006). The spectral analysis was applied for determination of mean spectrums (endmembers) using Spectral Angle Mapper (SAM), Spectral Feature Fitting (SFF), and Binary Encoding (BE) algorithms and USGS mineral spectral library as references spectra. To do this task, the reference library was resampled according to ASTER channels (figure 6). Finally the minerals of the spectral library having higher matching score to the endmembers based on the total scores of the three comparison methods were selected. The matching of absorption features on the reference and endmembers spectral profile were then visually checked to select best matches to the endmembers.The minerals Calcite, Montmorillonite, Tincalconite, Silica, and Hematite were therefore detected as representatives of 5 classes (table 1, figure 7). Table 1: assigned minerals to the detected classes. Class No Mineral type SAM SFF BE Class1 Calcite 0.902 0.704 0.778 Class2 Montmorillonite 0.865 0.843 0.889 Class3 Tincalconite 0.929 0.692 0.889 Class4 Quartz 0.814 0.763 0.778 Class5 Hematite 0.888 0.990 0.889 Figure 6: resampled montmorillonite and its original spectral profile. (a) (b) (c) (d) (e) Figure 7: absorption features of the mean spectrum of 5 classes (red lines) and their best fit from library (black li nes). (a) montmorillonite (b) calcite (c) tincalconite (d) quartz (e) hematite. After detection of endmembers, LSU algorithm was applied for computing of the abundances of detected minerals. The results were then integrated